Application of elastic net and infrared spectroscopy in the discrimination between defective and non-defective roasted coffees

AbstractThe quality of the coffee beverage is negatively affected by the presence of defective coffee beans and its evaluation still relies on highly subjective sensory panels. To tackle the problem of subjectivity, sophisticated analytical techniques have been developed and have been shown capable of discriminating defective from non-defective coffees after roasting. However, these techniques are not adequate for routine analysis, for they are laborious (sample preparation) and time consuming, and reliable, simpler and faster techniques need to be developed for such purpose. Thus, it was the aim of this study to evaluate the performance of infrared spectroscopic methods, namely FTIR and NIR, for the discrimination of roasted defective and non-defective coffees, employing a novel statistical approach. The classification models based on Elastic Net exhibited high percentage of correct classification, and the discriminant infrared spectra variables extracted provided a good interpretation of the models. The discrimination of defective and non-defective beans was associated with main chemical descriptors of coffee, such as carbohydrates, proteins/amino acids, lipids, caffeine and chlorogenic acids